ISSN 1000-1239 CN 11-1777/TP

Journal of Computer Research and Development ›› 2021, Vol. 58 ›› Issue (12): 2630-2644.doi: 10.7544/issn1000-1239.2021.20210997

Special Issue: 2021可解释智能学习方法及其应用专题

Previous Articles     Next Articles

Hierarchical Attention Network Based Interpretable Knowledge Tracing

Sun Jianwen1,2, Zhou Jianpeng1,2, Liu Sannüya1,2, He Feijuan3, Tang Yun4   

  1. 1(Faculty of Artificial Intelligence in Education, Central China Normal University, Wuhan 430079);2(National Engineering Laboratory for Educational Big Data(Central China Normal University), Wuhan 430079);3(Department of Computer, Xi’an Jiaotong University City College, Xi’an 710018);4(School of Psychology, Central China Normal University, Wuhan 430079)
  • Online:2021-12-01
  • Supported by: 
    This work was supported by the Major Program of National Science and Technology Innovation 2030 of China for New Generation of Artificial Intelligence (2020AAA0108804), the National Natural Science Foundation of China (62077021, 61977030, 61937001, 61807011), the Natural Science Basic Research Program of Shaanxi Province (2020JM-711), the Shaanxi Provincial Education Science Regulations “Thirteenth Five-Year” Plan Project (SGH20Y1397), the Special Research Project of Xi’an Jiaotong University City College (KCSZ01006), and the Teaching Reform Research Project for Postgraduates of Central China Normal University (2020JG14).

Abstract: Knowledge tracing is a data-driven learner modeling technology, which aims to predict learners’ knowledge mastery or future performance based on their historical learning data. Recently, with the support of deep learning algorithms, deep learning-based knowledge tracing has become a current research hotspot in the field. Aiming at the problems that deep learning-based knowledge tracing models generally have ‘black-box’ attributes, the decision-making process or results lack interpretability, and it is difficult to provide high-value education services such as learning attribution analysis and wrong cause backtracking, a Hierarchical Attention network based Knowledge Tracing model (HAKT) is proposed. By mining the multi-dimensional and in-depth semantic association between questions, a network structure containing three-layer attention of questions, semantics and elements is established, where graph attention neural network and self-attention mechanism are utilized for question representation learning, semantic fusion and questions retrieve. A regularization term to improve model interpretability is introduced into the loss function, with which a trade-off factor is incorporated to balance predictive performance and interpretability of model. Besides, we define an interpretability measurement index for the prediction results—fidelity, which can quantitatively evaluate the model interpretability. Finally, the experimental results on 6 benchmark datasets show that our method effectively improves the model interpretability.

Key words: knowledge tracing, interpretability, hierarchical attention, question semantic, fidelity

CLC Number: